Recent technological advances led to an appreciation of the genetic complexity of human acute myeloid leukemia (AML), but underlying progenitor cells remain poorly understood because their rarity precludes direct study. We developed a co-culture method integrating hypoxia, aryl hydrocarbon receptor inhibition and micro-environmental support via human endothelial cells to isolate these cells. X-chromosome inactivation studies of the least mature precursors derived following prolonged culture of CD34(+)/CD33(-) cells revealed polyclonal growth in highly curable AMLs, suggesting that mutations necessary for clonal expansion were acquired in more mature progenitors. Consistently, in core-binding factor (CBF) leukemias with known complementing mutations, immature precursors derived following prolonged culture of CD34(+)/CD33(-) cells harbored neither mutation or the CBF mutation alone, whereas more mature precursors often carried both mutations. These results were in contrast to those with leukemias with poor prognosis that showed clonal dominance in the least mature precursors. These data indicate heterogeneity among progenitors in human AML that may have prognostic and therapeutic implications.